GRDDL, RDFa, and DTDs, Re: Validation Update: success!

On Tue, 26 Jun 2007 20:10:29 +0100, Shane McCarron <shane@aptest.com>  
wrote:


>> So, if, philosophically, DTDs are syntactic rather than semantic, and
>> practically, invisible to javascript and XSLT, what is the advantage in
>> de-aligning RDFa from GRDDL by intending they be used to convey RDFa
>> semantics?
> (caveat - I know nothing of GRDDL)
>
I'm not an expert on GRDDL either, but I see it like this:
RDFa is a syntax for letting you express RDF in XHTML (and other XML  host  
dialects?).
GRDDL is a mechanism for specifying how your XHTML/XML document can be  
transformed into RDF.

What I want to understand better is how the two specifications (will) fit  
together.

> In order to add RDFa to XHTML (1.1) we need to extend the grammar.   
> There are some content model changes, extra attributes, etc.  In an  
> XHTML context, this means that you define a new markup language

So, from what you're saying, a DOCTYPE is a reasonable signifier of RDFa  
triples because RDFa is actually *part* of the markup language, rather  
than an *interpretation* of it. Is that right?

The qualm I have is that this doesn't seem totally compatible with GRDDL -  
GRDDL agents don't check DOCTYPEs, so say they ignore the DOCTYPE, they  
don't get all my RDFa triples. But if the agent is RDFa aware too, and  
checks the DOCTYPE as well as the profile, then there's a problem too;  
there's no way for me to say "this document doesn't contain RDFa" to the  
agent if I'm using the same DOCTYPE as a valid  RDFa document.

In the context of XHTML 1.0, this is a small annoyance - I can't use RDFa  
style attributes for any other purpose than RDFa *and* stay valid to the  
w3c validator.

But what about XHTML 2.0? You don't need a special RDFa to validate RDFa  
in XHTML 2.0 do you? (that was my perception at least) So will there be a  
separate DOCTYPE minted for XHTML 2.0+RDFa ? And any other XML dialect  
that can accommodate RDFa-style attributes? Or do GRDDL authors have to  
deliberately avoid anything that might look like RDFa ?

It seems to me it would be simpler and better by far (given the existence,  
and utility, of GRDDL) to use the GRDDL mechanism for signifying RDFa  
triples are present, and keep the DOCTYPEs for validation.

At the least, I think the presence of a @profile ought to override any  
other possible signifiers, such as DOCTYPEs.

What disadvantage is there in being consistent with GRDDL?

>
> You mention that you might need to change the rules for triples  
> generation.  I think that is incorrect.  The RDFa syntax document  
> should, when complete, define the rules for generating basic triples in  
> any context.

  At any rate, RDFa syntax isn't stable yet, but people are still using it,  
and the DOCTYPEs.

> These have to be immutable, IMHO

What I thought was, that if you do change the triple generation rules, you  
ought to change the signifier. If the DOCTYPE is the signifier, you have  
to provide a new DOCTYPE, even though the underlying XHTML syntax is the  
same - just the interpretation has changed. And you also have to change  
the profile uri. But if the profile is the dominant signifier, you only  
have to change the profile uri. That was my thought anyway. Not hugely  
important, but it still seems neater to have a distinction of DTD for XML  
syntax, and grddl profile for RDF semantics.


> As a document author, I need to be able to write some content once, and  
> KNOW it will be interpreted as I meant it to be by everyone who looks at  
> it.
>

Me too!

Yours,

Keith




-- 
Using Opera's revolutionary e-mail client: http://www.opera.com/mail/

Received on Wednesday, 27 June 2007 12:56:05 UTC